So here’s the question

Stillbirths have dropped by almost 8% in England since the smoking ban was introduced, research has shown.

The number of babies dying shortly after birth has also dropped by almost 8%, according to the study. A team led by Edinburgh University looked at information on more than 10 million births in England between 1995 and 2011.

OK. So, were the stillbirth and low weight numbers falling before the ban as well?

That is, was it the ban or did the ban simply coincide with something that was falling anyway?

25 thoughts on “So here’s the question”

  1. Full paper is available here, they use some sort of ‘counterfactual’ models to estimate what deaths etc they would’ve expected if the 2007 ban hadn’t occurred.

    Eyeballing the graphs for still birth looks particularly strange since, as with the ONS graph shown by Sarah, the previous fall seems to smoothly continue in reality whereas their ‘counterfactual’ seems to claim it would’ve have leveled out.

    http://www.nature.com/articles/srep13020

  2. And then I follow the Snowdon link and see he covered the same point I said 🙂

    Swear down, I googled the paper and spotted the ‘counterfactual’ dodge independently, just goes to show how obvious it is 😉

  3. As usual, non-scientists publish unscientific junk in a publication whose only connection to science is a word on its front cover.

    But the worrying thing is that this sort of attitude is dangerous. A friend of mine was having chemotherapy. In the initial consultation he was told that if he had chest pains while taking the medication, DIAL 999. On the front cover of a booklet he was given about his chemo was the warning that if he suffered chest pains DIAL 999. On the explanatory leaflet accompanying the drug it said if your suffer chest pains after taking this drug DIAL 999.

    Well, he did suffer chest pains and the first question he was asked was “Do you, or have you ever smoked?”
    “Yes, but I stopped 26 years ago.”
    “How many did you smoke a day?”
    “Wait a sec, isn’t this connected with my chemo?”
    “No, no no. Absolutely not. No.”
    “When do I restart the chemo then?”
    “Oh, no , no more chemo for you, it’s radiotherapy now”

    Comforting, eh?

  4. Doctors – GPs mostly – are doing themselves no favors at all with the smoking interrogation malarkey.

    I’ve lost count of the number of acquaintances who’ve remarked on the subject after a visit to a GP – and not approvingly.

    The removal of UK Public ‘Elf from NHS to local government – i.e. clinicians shoving the cuckoos out of the nest tells you much about what these parasites are.

    I believe it would be useful to put a name to each and every PH utterance like this revelation – since the perps. are serial offenders who rely on anonymity……

  5. Can any statistician here interpret the “statistically significant” claims?

    I only ask because eye-balling the graphs – pre and post 2007 – and taking into account the data ranges (high low) involved, as others have observed, doesn’t intuitively look that conclusive (ie, I would presume quite chunky error margins from eye-balling)?

    Second, what happened around 1992? That change in trend looks far more interesting / decisive?

  6. “what happened around 1992?” Major beat Kinnock and so the Labour Terror was deferred for five years.

  7. “Major beat Kinnock and so the Labour Terror was deferred for five years.”

    Absolutely!

    And yet, with hindsight, had Labour narrowly won, then:

    Kinnock, exit from ERM, economic ineptitude, less focus on Tory sleaze / infighting over those 5 years, etc, the chances are good that we might have had just the one term of terror.

    Just imagine – Tony Blair, Gordon Brown and David Cameron might simply never have happened…

  8. were the stillbirth and low weight numbers falling before the ban as well

    Yes they were.

    was it the ban or did the ban simply coincide with something that was falling anyway?

    The analysis explored various models, all incorporating underlying trends in the data.

    It’s fair to conclude that stillbirths and neonatal deaths fell after July 2007 more than would have been expected from previous trends. Whether the smoking ban was causative one cannot say.

    Incidentally, quite a lot of what looks like noise in the charts is seasonal variation. The case might look a lot clearer if you took that out.

  9. “Incidentally, quite a lot of what looks like noise in the charts is seasonal variation. The case might look a lot clearer if you took that out.”

    Some of it, you’re absolutely right – I didn’t spot that. And which would narrow the error bars.

  10. bloke (not) in spain

    I think we can assume the trend is absolutely inconclusive because the Graun neither publishes the graph nor links to it. Just obsesses over old smoking articles.
    You can be sure it would be there if it was.
    Like sussing all con-tricks, it’s what you aren’t getting to see tells where the con is.

  11. Relative % used not absolute to make the number look fantastic. What % of births are still born and what is 8% of that?

  12. <donning the stats geek outfit>
    @tlibt1, Rob: Counterfactual models are well established in medical and epidemiological models. Ignorance of or lack of understanding of a concept is not a sound basis for rejection of said concept.

    @PF: On the significance bit: the p-values they quote can be thought of as the probability the result is a fluke. So low values are good, 0.05 or lower is nice when you want something published.

    In terms of methodology they seem to have done a pretty good job. The approach they describe for developing the models is sound. I would have liked to see the full details of the models they have fitted in the appendix, and a reference to the compiled data – then I could I have run the numbers with about as much effort as I would like to on a weekend.

    @Dounguan John: Presumably it is the 7.8% reduction they are referring to – It is actually a reduction of the odds that the paper is reporting. True to their usual grasp of numbers and measures the G report it as something completely else, totally distorting any sensible discussion.

  13. anon

    Thanks for the ponter / steer on the 95% CL’s, I understand how they are describing that now.

    btw, do you know what happened around 1992. That looks to be a much more significant change in trend (eye-balling)?

  14. @PF: 1992 – I don’t have a clue. I am leaning towards acabab’s suggestion – someone changed the measure directly or indirectly. If you ever want a statistician to start swearing aloud, using some pretty imaginative profanities, suggest changing well established measures.

    CI’s are a misunderstood beast, almost as badly as p-values. Actually, they are probably even worse understood. Wikipedia has a good summary on meaning and interpretation, and it shows why people often get it wrong.

Leave a Reply

Your email address will not be published. Required fields are marked *